PubMed Health⌕ Search

Biomedical subjects

M S Friedrichs

Publications and source records attributed to M S Friedrichs.

17 recordsLinked to original sources
PubMed1998

Backbone and side chain dynamics of uncomplexed human adipocyte and muscle fatty acid-binding proteins.

Adipocyte lipid-binding protein (A-LBP) and muscle fatty acid-binding protein (M-FABP) are members of a family of small ( approximately 15 kDa) cytosolic proteins that are involved in the metabolism of fatty acids and other lipid-soluble molecules. Although highly homologous (65%) and structurally very similar, A-LBP and M-FABP display distinct ligand binding characteristics. Since ligand binding may be influenced by intrinsic protein dynamical properties, we have characterized the backbone and side chain dynamics of uncomplexed (apo) human A-LBP and M-FABP. Backbone dynamics were characterized by measurements of 15N T1 and T2 values and ¿1H¿-15N NOEs. These data were analyzed using model-free spectral density functions and reduced spectral density mapping. The dynamics of methyl-containing side chains were charaterized by measurements of 2H T1 and T1rho relaxation times of 13C1H22H groups. The 2H relaxation data were analyzed using the model-free approach. For A-LBP, 15N relaxation data were obtained for 111 residues and 2H relaxation data were obtained for 42 methyl groups. For M-FABP, 15N relaxation data were obtained for 111 residues and 2H relaxation data were obtained for 53 methyl groups. The intrinsic flexibilities of these two proteins are compared, with particular emphasis placed on binding pocket residues. There are a number of distinct dynamical differences among corresponding residues between the two proteins. In particular, many residues display greater backbone picosecond to nanosecond and/or microsecond to millisecond time scale mobility in A-LBP relative to M-FABP, including F57, K58, and most residues in alpha-helix 2 (residues 28-35). Variations in the dynamics of this region may play a role in ligand selectivity. The side chains lining the fatty acid binding pocket display a wide range of motional restriction in both proteins. Side chains showing distinct dynamical differences between the two proteins include those of residues 20, 29, and 51. This information provides a necessary benchmark for determining dynamical changes induced by ligand binding and may ultimately lead to an enhanced understanding of ligand affinity and selectivity among fatty acid-binding proteins.

Adipocytes↗
PubMed1997

Characterization of NADP+ binding to perdeuterated MurB: backbone atom NMR assignments and chemical-shift changes.

Backbone-atom resonances have been assigned for both the substrate-free and the NADP+-complexed forms of UDP-N-acetylenolpyruvylglucosamine reductase (MurB), a monomeric, 347-residue (38.5 kDa) flavoenzyme essential for bacterial cell-wall biosynthesis. NMR studies were performed using perdeuterated, uniformly 13C/15N-labeled samples of MurB. In the case of substrate-free MurB, one or more backbone atoms have been assigned for 334 residues (96%). The assigned backbone atoms include 309 1HN and 15N atoms (94%), 315 13CO atoms (91%), 331 13C(alpha) atoms (95%), and 297 13C(beta) atoms (93%). For NADP+-complexed MurB, one or more backbone atoms have been assigned for 313 residues (90%); these include 283 1HN and 15N atoms (86%), 305 13CO atoms (88%), 310 13C(alpha) atoms (89%), and 269 13C(beta) atoms (84%). The strategies used for obtaining resonance assignments are described in detail. Information on the secondary structure in solution for both the substrate-free and NADP+-complexed forms of the enzyme has been derived both from 13C(alpha) and 13C(beta) chemical-shift deviations from random-coil values and from 1HN-1HN NOEs. These data are compared to X-ray crystallographic structures of substrate-free MurB and MurB complexed with the UDP-N-acetylglucosamine enolpyruvate (UNAGEP) substrate. NADP+ binding induces significant chemical-shift changes in residues both within the known UNAGEP and FAD binding pockets and within regions known to undergo conformational changes upon UNAGEP binding. The NMR data indicate that NADP+ and UNAGEP utilize the same binding pocket and, furthermore, that the binding of NADP+ induces structural changes in MurB. Finally, many of the residues within the UNAGEP/NADP+ binding pocket were difficult to assign due to dynamic processes which weaken and/or broaden the respective resonances. Overall, our results are consistent with MurB having a flexible active site.

Amino Acid Sequence↗
PubMed1997

Solution structure of the Grb2 N-terminal SH3 domain complexed with a ten-residue peptide derived from SOS: direct refinement against NOEs, J-couplings and 1H and 13C chemical shifts.

Refined ensembles of solution structures have been calculated for the N-terminal SH3 domain of Grb2 (N-SH3) complexed with the ac-VPPPVPPRRR-nh2 peptide derived from residues 1135 to 1144 of the mouse SOS-1 sequence. NMR spectra obtained from different combinations of both 13C-15N-labeled and unlabeled N-SH3 and SOS peptide fragment were used to obtain stereo-assignments for pro-chiral groups of the peptide, angle restraints via heteronuclear coupling constants, and complete 1H, 13C, and 15N resonance assignments for both molecules. One ensemble of structures was calculated using conventional methods while a second ensemble was generated by including additional direct refinements against both 1H and 13C(alpha)/13C(beta) chemical shifts. In both ensembles, the protein:peptide interface is highly resolved, reflecting the inclusion of 110 inter-molecular nuclear Overhauser enhancement (NOE) distance restraints. The first and second peptide-binding sub-sites of N-SH3 interact with structurally well-defined portions of the peptide. These interactions include hydrogen bonds and extensive hydrophobic contacts. In the third highly acidic sub-site, the conformation of the peptide Arg8 side-chain is partially ordered by a set of NOE restraints to the Trp36 ring protons. Overall, several lines of evidence point to dynamical averaging of peptide and N-SH3 side-chain conformations in the third subsite. These conformations are characterized by transient charge stabilized hydrogen bond interactions between the peptide arginine side-chain hydrogen bond donors and either single, or possibly multiple, acceptor(s) in the third peptide-binding sub-site.

Adaptor Proteins, Signal Transducing↗
PubMed1995

Refined solution structure of human profilin I.

Profilin is a ubiquitous eukaryotic protein that binds to both cytosolic actin and the phospholipid phosphatidylinositol-4,5-bisphosphate. These dual competitive binding capabilities of profilin suggest that profilin serves as a link between the phosphatidyl inositol cycle and actin polymerization, and thus profilin may be an essential component in the signaling pathway leading to cytoskeletal rearrangement. The refined three-dimensional solution structure of human profilin I has been determined using multidimensional heteronuclear NMR spectroscopy. Twenty structures were selected to represent the solution conformational ensemble. This ensemble of structures has root-mean-square distance deviations from the mean structure of 0.58 A for the backbone atoms and 0.98 A for all non-hydrogen atoms. Comparison of the solution structure of human profilin to the crystal structure of bovine profilin reveals that, although profilin adopts essentially identical conformations in both states, the solution structure is more compact than the crystal structure. Interestingly, the regions that show the most structural diversity are located at or near the actin-binding site of profilin. We suggest that structural differences are reflective of dynamical properties of profilin that facilitate favorable interactions with actin. The global folding pattern of human profilin also closely resembles that of Acanthamoeba profilin I, reflective of the 22% sequence identity and approximately 45% sequence similarity between these two proteins.

Actins↗
PubMed1995

High-resolution solution structure of siamycin II: novel amphipathic character of a 21-residue peptide that inhibits HIV fusion.

The 21-amino acid peptides siamycin II (BMY-29303) and siamycin I (BMY-29304), derived from Streptomyces strains AA3891 and AA6532, respectively, have been found to inhibit HIV-1 fusion and viral replication in cell culture. The primary sequence of siamycin II is CLGIGSCNDFAGCGYAIVCFW. Siamycin I differs by only one amino acid; it has a valine residue at position 4. In both peptides, disulfide bonds link Cys1 with Cys13 and Cys7 with Cys19, and the side chain of Asp9 forms an amide bond with the N-terminus. Siamycin II, when dissolved in a 50:50 mixture of DMSO and H2O, yields NOESY spectra with exceptional numbers of cross peaks for a peptide of this size. We have used 335 NOE distance constraints and 13 dihedral angle constraints to generate an ensemble of 30 siamycin II structures; these have average backbone atom and all heavy atom rmsd values to the mean coordinates of 0.24 and 0.52 A, respectively. The peptide displays an unusual wedge-shaped structure, with one face being predominantly hydrophobic and the other being predominantly hydrophilic. Chemical shift and NOE data show that the siamycin I structure is essentially identical to siamycin II. These peptides may act by preventing oligomerization of the HIV transmembrane glycoprotein gp41, or by interfering with interactions between gp41 and the envelope glycoprotein gp120, the cell membrane or membrane-bound proteins [Frèchet, D. et al. (1994) Biochemistry, 33, 42-50]. The amphipathic nature of siamycin II and siamycin I suggests that a polar (or apolar) site on the target protein may be masked by the apolar (or polar) face of the peptide upon peptide/protein complexation.

Amino Acid Sequence↗
PubMed1994

Sequential 1H, 13C, and 15N NMR assignments and solution conformation of apokedarcidin.

Kedarcidin is a recently discovered antitumor antibiotic chromoprotein. The solution conformation of the kedarcidin apoprotein (114 residues) has been characterized by heteronuclear multidimensional NMR spectroscopy. Sequence-specific backbone atom resonance assignments were obtained for a uniformly 13C/15N-enriched sample of apokedarcidin via a semiautomated analysis of 3D HNCACB, 3D CBCA-(CO)NH, 4D HNCAHA, 4D HN(CO)CAHA, 3D HBHA(CO)NH, and 3D HNHA(Gly) spectra. Side-chain assignments were subsequently obtained by analysis of (primarily) 3D HCCH-TOCSY and HCCH-COSY spectra. A qualitative analysis of the secondary structure is presented on the basis of 3J alpha NH coupling constants, deviations of 13C alpha and 13C beta chemical shifts from random coil values, and NOEs observed in 3D 15N- and 13C-edited NOESY-HSQC spectra. This analysis revealed a four-stranded antiparallel beta-sheet, a three-stranded antiparallel beta-sheet, and two two-standed antiparallel beta-sheets. The assignments of cross-peaks in the 3D NOESY spectra were assisted by reference to a preliminary model of apokedarcidin built using the program CONGEN starting from the X-ray structure of the homologous protein aponeocarzinostatin. An ensemble of 15 apokedarcidin solution structures has been generated by variable target function minimization (DIANA program) and refined by simulated annealing (X-PLOR program). The average backbone atom root-mean-square difference between the individual structures and the mean coordinates is 0.68 +/- 0.08 A. The overall fold of apokedarcidin is well-defined; it is composed of an immunoglobulin-like seven-stranded antiparallel beta-barrel and a subdomain containing two antiparallel beta-ribbons. Highly similar tertiary structures have been previously reported for the related proteins neocarzinostatin, macromomycin, and actinoxanthin. Important structural features are revealed, including the dimensions of the chromophore-binding pocket and the locations of side chains that are likely to be involved in chromophore stabilization.

Amino Acid Sequence↗
PubMed1994

Solution structure of an isolated antibody VL domain.

The solution structure of the isolated VL domain of the anti-digoxin antibody 26-10 has been determined using data derived from heteronuclear multi-dimensional nuclear magnetic resonance (n.m.r.) experiments. Analytical ultracentrifugation and n.m.r. data demonstrate that the VL domain is only weakly associating (Kd = 2.5 (+/- 0.7) mM) and that it experiences a rapid monomer/dimer equilibrium under the n.m.r. experimental conditions. Therefore, the results reported here represent the first structure determination of an antibody VL domain in the absence of fixed quaternary interactions. The structure determination is based on 930 proton-proton distance constraints, 113 dihedral angle constraints, and 46 hydrogen bond constraints. Eighty initial structures were calculated with the variable target function program DIANA; of these, 31 were accepted on the basis of satisfaction of constraints (no distance constraint violations > 0.5 A; target function < 3.0 A2). Accepted DIANA structures were refined by restrained energy minimization using the X-PLOR program. The 15 best energy-minimized DIANA structures were chosen as a representative ensemble of solution conformations. The average root-mean-square differences (r.m.s.d.) between the individual structures of this ensemble and the mean coordinates is 0.85 (+/- 0.10) A for all backbone atoms and 1.29 (+/- 0.10) A for all heavy atoms. For beta-strands A, B, C, D, E and F, the average backbone atom r.m.s.d. to the mean structure is 0.46 (+/- 0.06) A. A higher-resolution ensemble, with all backbone atom and all heavy atom r.m.s.d.s. to the mean coordinates of 0.54 (+/- 0.08) A and 0.98 (+/- 0.12) A, respectively, was obtained by X-PLOR simulated annealing refinement of the 15 energy-minimized DIANA structures. A detailed analysis of the original ensemble of 15 energy-minimized DIANA structures is presented, as this ensemble retains a broader, and possibly more realistic, sampling of conformation space. The backbone atom and all heavy atom r.m.s.d.s between the mean energy-minimized DIANA structure and the X-ray derived coordinates of the VL domain within the Fab/digoxin complex are 1.05 A and 1.56 A, respectively. Subtle differences between the solution and X-ray structures occur primarily in CDR2, CDR3, beta-strands A, F and G, and localized regions of hydrophobic packing. Overall, these results demonstrate that the 26-10 VL domain conformation is determined primarily by intradomain interactions, and that quaternary VL-VH association induces relatively minor conformational adjustments.

Computer Graphics↗
PubMed1994

An automated procedure for the assignment of protein 1HN, 15N, 13C alpha, 1H alpha, 13C beta and 1H beta resonances.

A computer algorithm that determines the 1HN, 15N, 13C alpha, 1H alpha, 13C beta and 1H beta chemical-shift assignments of protein residues with minimal human intervention is described. The algorithm is implemented as a suite of macros that run under a modified version of the FELIX 1.0 program (Hare Research, Bothell, WA). The input to the algorithm is obtained from six multidimensional, triple-resonance experiments: 3D HNCACB, 3D CBCA(CO)HN, 4D HNCAHA, 4D HN(CO)CAHA, 3D HBHA(CO)NH and 3D HNHA(Gly). For small proteins, the two 4D spectra can be replaced by either the 3D HN(CA)HA, 3D H(CA)NNH, or the 15N-edited TOCSY-HSQC experiments. The algorithm begins by identifying and collecting the intraresidue and sequential resonances of the backbone and 13C beta atoms into groups. These groups are sequentially linked and then assigned to residues by matching the 13C alpha and 13C beta chemical-shift profiles of the linked groups to that of the protein's primary structure. A major strength of the algorithm is its ability to overcome imperfect data, e.g., missing or overlapping peaks. The viability of the procedure is demonstrated with two test cases. In the first, NMR data from the six experiments listed above were used to reassign the backbone resonances of the 93-residue human hnRNP C RNA-binding domain. In the second, a simulated cross-peak list, generated from the published NMR assignments of calmodulin, was used to test the ability of the algorithm to assign the backbone resonances of proteins containing internally homologous segments. Finally, the automated method was used to assign the backbone resonances of apokedarcidin, a previously unassigned, 114-residue protein.

Algorithms↗
PubMed1993

Relaxation study of the backbone dynamics of human profilin by two-dimensional 1H-15N NMR.

The dynamic properties of 111 backbone HN sites in uncomplexed human profilin, a protein of 139 residues, have been characterized by two-dimensional inverse-detected 1H-15N NMR spectroscopy. Heteronuclear (1H)-15N nuclear Overhauser effects and 15N longitudinal and transverse relaxation rates have been analyzed in terms of model-free spectral density functions and exchange contributions to transverse relaxation rates. Relatively high mobilities on the nanosecond time-scale are observed for Asp26 and Ser27, which form part of a loop connecting beta-strands A and B, and for Thr92 through Ala95, which are in a loop connecting beta-strands E and F. Significant exchange contributions, indicative of motions on the microsecond to millisecond time-scale, have been obtained for 30 residues. These include Leu77, Asp80 and Gly81 of a loop between beta-strands D and E, Ser84 and Met85 of beta-strand E, Gly121 of a loop connecting beta-strand G and the C-terminal helix, and Gln138, which is next to the C-terminal residue Tyr139. Some of the regions showing high flexibility in profilin are known to be involved in poly-L-proline binding.

Amino Acid Sequence↗
PubMed1993

Characterization of the three-dimensional solution structure of human profilin: 1H, 13C, and 15N NMR assignments and global folding pattern.

Human profilin is a 15-kDa protein that plays a major role in the signaling pathway leading to cytoskeletal rearrangement. Essentially complete assignment of the 1H, 13C, and 15N resonances of human profilin have been made by analysis of multidimensional, double- and triple-resonance nuclear magnetic resonance (NMR) experiments. The deviation of the 13C alpha and 13C beta chemical shifts from their respective random coil values were analyzed and correlate well with the secondary structure determined from the NMR data. Twenty structures of human profilin were refined in the program X-PLOR using a total of 1186 experimentally derived conformational restraints. The structures converged to a root mean squared distance deviation of 1.5 A for the backbone atoms. The resultant conformational ensemble indicates that human profilin is an alpha/beta protein comprised of a seven-stranded, antiparallel beta-sheet and three helices. The secondary structure elements for human profilin are quite similar to those found in Acanthamoeba profilin I [Archer, S. J., Vinson, V. K., Pollard, T. D., & Torchia, D. A. (1993), Biochemistry 32, 6680-6687], suggesting that the three-dimensional structure of Acanthamoeba profilin I should be analogous to that determined here for human profilin. The structure determination of human profilin has facilitated the sequence alignment of lower eukaryotic and human profilins and provides a framework upon which the various functionalities of profilin can be explored. At least one element of the actin-binding region of human profilin is an alpha-helix. Two mechanisms by which phosphatidylinositol 4,5-bisphosphate can interfere with actin-binding by human profilin are proposed.

Amino Acid Sequence↗
PubMed1993

Micelle-bound conformational preferences of a peptide derived from a murine major histocompatibility complex class I molecule.

Models of the micelle-bound conformation of a 17-residue major histocompatibility complex-derived peptide, [Ala85]Dk(69-85), have been determined by NMR spectroscopy and simulated annealing calculations. This peptide is a truncated, substituted version of Dk(61-85), which is a fragment of the murine major histocompatibility complex class I molecule H-2Dk. Dk(61-85) has been shown to adopt an ordered conformation required for augmentation of insulin-stimulated glucose uptake (Stagsted, J., Baase, W. A., Goldstein, A., and Olsson, L. (1991) J. Biol. Chem. 266, 12844-12847). [Ala85]Dk(69-85) retains full biological activity. Thirty-eight converged NMR structures of [Ala85]Dk(69-85) bound to dodecyl phosphocholine micelles have been generated. The NMR-derived models display a propensity for a type-I beta-bend involving residues 73-76 and an amphipathic helical region involving residues 77-84. CD spectra yield a helical content (8% at 20-25 degrees C) consistent with transient, partial helix formation. The relative orientation of the beta-bend region with respect to the helical region is not well defined by the NMR data. This may reflect true heterogeneity of the micelle-bound conformation. The NMR structures were compared with a model of [Ala85]Dk(69-85) derived from the x-ray coordinates of the human major histocompatibility complex class I allele HLA-Aw68 (Garrett, T. P. J., Saper, M. A., Bjorkmann, P. J., Strominger, T. L., and Wiley, D. C. (1989) Nature 342, 692-696). Structural features that are important for the bioactivity of [Ala85]Dk(69-85) are discussed with reference to reported structure-activity relationships (Stagsted, J., Mapelli, C., Myers, C., Matthews, B. W., Anfinsen, C. B., Goldstein, A., and Olsson, L. (1993) Proc. Natl. Acad. Sci. U.S.A., in press). A general description of the structural properties of the putative receptor site(s) that are likely to be required for binding [Ala85]Dk(69-85) is given.

Amino Acid Sequence↗
PubMed1993

Characterization of the backbone dynamics of an anti-digoxin antibody VL domain by inverse detected 1H-15N NMR: comparisons with X-ray data for the Fab.

The dynamic behavior of the polypeptide backbone of a recombinant antidigoxin antibody VL domain has been characterized by measurements of 15NT1 and T2 relaxation times, 1H-15N NOE values, and 1H-2H exchange rates. These data were acquired with 2D inverse detected heteronuclear 1H-15N NMR methods. The relaxation data are interpreted in terms of model free spectral density functions and exchange contributions to transverse relaxation rates R2 (= 1/T2). All characterized residues display low-amplitude picosecond time-scale librational motions. Fifteen residues undergo conformational changes on the nanosecond timescale, and 24 residues have significant R2 exchange contributions, which reflect motions on the microsecond to millisecond time-scale. For several residues, microsecond to millisecond motions of nearby aromatic rings are postulated to account for some or all of their observed R2 exchange contributions. The measured 1H-2H exchange rates are correlated with hydrogen bonding patterns and distances from the solvent accessible surface. The degree of local flexibility indicated by the NMR measurements is compared to crystallographic B-factors derived from X-ray analyses of the native Fab and the Fab/digoxin complex. In general, both the NMR and X-ray data indicate enhanced flexibility in the turns, hypervariable loops, and portions of beta-strands A, B, and G. However, on a residue-specific level, correlations among the various NMR data, and between the NMR and X-ray data, are often absent. This is attributed to the different dynamic processes and environments that influence the various observables. The combined data indicate that certain regions of the VL domain, including the three hypervariable loops, undergo dynamic changes upon VL:VH association and/or complexation with digoxin. Overall, the 26-10 VL domain exhibits relatively low flexibility on the ps-ns timescale. The possible functional consequences of this result are considered.

Digoxin↗
PubMed1993

Aliphatic 1H and 13C resonance assignments for the 26-10 antibody VL domain derived from heteronuclear multidimensional NMR spectroscopy.

Extensive 1H and 13C assignments have been obtained for the aliphatic resonances of a uniformly 13C- and 15N-labeled recombinant VL domain from the anti-digoxin antibody 26-10. Four-dimensional triple resonance NMR data acquired with the HNCAHA and HN(CO)CAHA pulse sequences [Kay et al. (1992) J. Magn. Reson., 98, 443-450] afforded assignments for the backbone HN, N, H alpha and C alpha resonances. These data confirm and extend HN, N and H alpha assignments derived previously from three-dimensional 1H-15N NMR studies of uniformly 15N-labeled VL domain [Constantine et al. (1992), Biochemistry, 31, 5033-5043]. The identified H alpha and C alpha resonances provided a starting point for assigning the side-chain aliphatic 1H and 13C resonances using three-dimensional HCCH-COSY and HCCH-TOCSY experiments [Clore et al. (1990), Biochemistry, 29, 8172-8184]. The C alpha and C beta chemical shifts are correlated with the VL domain secondary structure. The extensive set of side-chain assignments obtained will allow a detailed comparison to be made between the solution structure of the isolated VL domain and the X-ray structure of the VL domain within the 26-10 Fab.

Amino Acid Sequence↗
PubMed1992

Determination of the three-dimensional solution structure of ragweed allergen Amb t V by nuclear magnetic resonance spectroscopy.

Analysis of two-dimensional NMR experiments has afforded essentially complete assignment of all proton resonances in the allergenic protein Amb t V. Conformational constraints were obtained from the NMR data in three forms: interproton distances derived from NOE cross-peak intensities of NOESY spectra, torsion angle constraints derived from J-coupling constants of COSY and PE-COSY spectra, and hydrogen bond constraints derived from hydrogen-exchange experiments. Conformations of Amb t V with low constraint violations were generated using dynamic simulated annealing in the program XPLOR. The refined structures are comprised of a C-terminal alpha-helix, a short stretch of triple-stranded antiparallel beta-sheet, and several loops. In addition, the cystine partners of the four disulfide linkages (for which there are no biochemical data) have been assigned. The refined structures of Amb t V will allow us to suggest surface substructures for the Amb V allergens that are likely to participate in B cell epitopes and will assist us in defining the Ia/T cell epitopes that interact with the MHC class II (or Ia) molecule and the T cell receptor leading to the induction of the immune response to Amb t V.

Allergens↗
PubMed1991

Generalized protein tertiary structure recognition using associative memory Hamiltonians.

In previous papers, a method of protein tertiary structure recognition was described based on the construction of an associative memory Hamiltonian, which encoded the amino acid sequence and the C alpha co-ordinates of a set of database proteins. Using molecular dynamics with simulated annealing, the ability of the Hamiltonian to successfully recall the structure of a protein in the memory database was successfully demonstrated, as long as the total number of database proteins did not exceed a characteristic value, called the capacity of the Hamiltonian, equal to 0.5N to 0.7N, where N is the number of amino acid residues in the protein to be recalled. In this paper, we describe the development of additional methods to increase the capacity of the Hamiltonian, including use of a more complete representation of the protein backbone and the incorporation of contextual information into the Hamiltonian through the use of secondary structure prediction. In addition, we further extend the ability of associative memory models to predict the tertiary structures of proteins not present in the protein data set, by making the Hamiltonian invariant with respect to biological symmetries that represent site mutations and insertions and deletions. The ability of the Hamiltonian to generalize from homologous proteins to an unknown protein in the presence of other unrelated proteins in the data set is demonstrated.

Cytochromes↗